Short answers, and where the long ones live
Open a question to read the answer. Each one links to the page that covers it in full.
What this is
What does this site do?
An option price before an earnings report contains two separate things: the stock’s ordinary day-to-day movement, and the jump the report itself is expected to cause. We separate them, and rank the report’s own price against that same company’s past reports.
For every report that has already happened we publish what the options market implied beforehand, what the stock did, and how far implied volatility collapsed afterwards.
Is this trading advice, or a signal?
Neither. It is a measuring instrument. It reports what things cost and what has happened before, and it makes no recommendation about what to trade or which way a stock will go.
The one forward-looking number we publish is a prediction about implied volatility, not about which way a stock will move, and we score it in public so you can check it.
Did you invent this method?
No. The way we separate the report’s own jump from a stock’s ordinary movement is a published estimator with a peer-reviewed literature behind it. The three papers that define it are linked below.
What is ours is the implementation, the measurement, and the scored prediction record. So is one deliberate departure: we treat the event as the whole reporting session rather than an instantaneous jump, because that is what our realized-move numbers are measured over. The methodology page says where each paper and our version part company, and why.
Who is it for?
Someone who already trades around earnings and is currently assembling these numbers by hand. It replaces that spreadsheet work with a page.
It is not a starting point for someone new to options. It tells you the temperature accurately and says nothing about whether to go outside.
Is the data live?
No. Every number is sampled at fixed points around a report (the afternoon before and the session after) and published that evening. Nothing updates during the trading day.
So it is a tool for planning and reviewing trades. A company reporting after the close is reviewable two mornings later; one reporting before the open, the next morning. All data is delayed.
The numbers
What is the “expected move”?
It is what the at-the-money options in the expiry containing the report imply the stock will move. We publish the move implied before each past report beside the move that followed.
We publish two versions: the raw one the whole expiry implies, and the split one, with the stock’s ordinary movement taken out. The second is the one the board ranks on.
Why split the move into two parts?
Because the raw number mostly tells you how jumpy the stock is. A quiet company facing an uncertain report and a habitually wild company facing a dull one can price the same expected move for very different reasons.
Separating the report’s own jump from the stock’s ordinary movement makes the ranking comparable, and lets us predict the collapse that follows.
What do “rich” and “cheap” mean, and does rich mean the trade wins?
Rich and cheap place the report’s own priced move, with ordinary movement stripped out, as a percentile of the same company’s past reports. A name at the top of the board is expensive for itself; the ranking does not compare it with other companies.
No. The ranking looks only at what the options cost, never at what the stock did next, and reports at the same percentile have had very different outcomes. Options are usually somewhat expensive before earnings, because sellers charge for the risk.
Why is your IV crush number bigger than the one I see elsewhere?
Because we measure it on the expiry that contains the report. A 30-day constant-maturity measure blends that expiry with expiries holding no event, so it reports a systematically shallower collapse than the one a position is in.
We store and display the 30-day number beside ours so the comparison is auditable. The time of day moves the figure too. Our headline runs from the afternoon before the report to the same time the session after, which captures the opening collapse plus a full session of drift. A figure measured at the open is a different, usually smaller, number, and we publish that one separately.
Why measure the realized move at 10:00 as well as at the close?
Close-to-close is the conventional measure and it prices an exit at 16:00 the day after the report. A position is usually closed nearer the open.
So we publish the move to the opening auction and to the 09:55–10:05 volume-weighted price as well, and how often the move extended between them. The three anchors never stand in for one another: if one is missing, it stays missing.
What is IV rank here?
Where a name’s current 30-day implied volatility sits inside its own trailing year, on a 0–100 scale. It says whether options on this name are expensive for it right now, regardless of any upcoming report.
What is the short-straddle outcome table?
One fixed mechanical question asked backwards over a name’s stored history: sell the at-the-money straddle the afternoon before the report and buy it back the session after, at the open, at 10:00 and at the close.
Nothing about it is configurable, so it is a fixed measurement, not a strategy or a backtest. It pays the bid-ask spread both ways; the version without the spread is shown beside it to show what the spread costs.
How well it works
How accurate are your predictions?
The accuracy record has the current figures. It is public, needs no account, and keeps every result, bad days included.
It reports the average miss (mean absolute error) and the signed bias, in points of the collapse, alongside the share of reports we predicted at all. That share is published because it is the number that can be gamed: an error rate over only the reports we chose to predict would look better than it is.
Why are there two records, and why are they never added together?
One is a forecasting record: predictions frozen the session before each report and scored afterwards. The other is the same method re-run over history we already had. Only the first can be evidence that the method predicts.
A blended figure would describe neither, so the site never combines them. Frozen predictions cannot be rewritten afterwards; the database itself refuses the change.
Do you beat simply guessing?
The accuracy page scores our frozen prediction against three simple guesses on the same reports: no collapse at all, every report collapsing by the average, and each company repeating its own past average.
Are there biases you know about?
Yes, and we publish them. The textbook prediction overshoots how far implied volatility falls, and more so when a Fed decision or an election is still ahead. We correct most of that with numbers refitted each quarter on reports already graded, and publish the signed bias that is left as well as the absolute error. The decomposition assumes a normally distributed event day, which real reports are not; we have measured what that costs and where it lands.
The accuracy record also covers only companies liquid enough for the decomposition to be trusted. Any filter on an accuracy record changes it, so we say which way this one cuts: it makes both the error and the coverage look better.
Data and coverage
Why is this cell blank?
We leave a cell blank when we could not compute the number reliably. Every blank carries a marker saying why: thin quotes, a decomposition band too wide to rank on, or intraday data missing from the tape.
We tell a gap in the market data apart from an outage of ours, and we publish the share of reports we could measure instead of averaging over whatever is left.
What do the small superscript markers next to a number mean?
Each marker is a caveat recorded when the number was computed. We show them so you can judge the number yourself. On a computer, hover over one to see what it means; the methodology page explains each kind.
A marked number is still shown. Usually a marker affects only whether the row is ranked.
Where does the data come from?
Option chains, implied volatilities and greeks are licensed from a third-party market data provider; intraday equity and option bars from a second; the forward earnings calendar is sourced separately and cross-checked against the record of past reports.
Nothing displayed is a vendor’s pre-computed field. Every number on the site is computed here from raw chains and price bars by the published formulas, which is what makes the cross-checks meaningful.
Which companies do you cover?
A company gets a public history page once it has at least four past reports we are confident in, because a page with a noisy expected move is worse than no page. Those pages are free and need no account.
The forward board lists every company we cover that reports in the next two weeks, but ranks only those whose options give a clean reading. Most of those are the companies liquid enough for the decomposition to be worth trusting, a list re-measured monthly from traded liquidity. The rest show their numbers unranked, or sit in the held-out list with the reason.
Why is the board sometimes nearly empty?
Because earnings arrive in seasons. The board lists companies reporting in the next two weeks, and between seasons very few report, so an empty stretch in late September or late December is the calendar, not a fault.
The off-season board covers those weeks. It ranks the names we read every night, mostly the more heavily traded companies and the ETFs, by where each one’s 30-day implied volatility sits between its low and its high of the past year. Names reporting in the next two weeks are listed apart, because their implied volatility carries the report.
When does the site update?
Once nightly, after the session closes. The bar at the top of every data page shows the last session we hold data through and whether the most recent run completed, so you can see at a glance how fresh the data is.
The board is read once a day. Is yesterday’s close good enough?
For ranking reports, it holds up. The board compares each company’s report move with its own past reports, and that move changes slowly over the sessions before a report.
We tested it on past reports by ranking each one on the close a session earlier, the reading you would have traded on. The reports ranked rich still moved less than they were priced to, and the cheap ones moved more, by about the same margin as on the last reading before the report. Ranks do shift overnight: some reports change band, almost always to the next one over. Treat a name near the edge of a band as a close call.
I think one of your numbers is wrong.
Please tell us. Every history page has a “Report a data issue” link at the foot that fills in the company and date for you. We would rather hear about a wrong number than have you decide the numbers are unreliable without telling us.
Where a number turns out to be wrong we correct the underlying record and recompute that report. Frozen predictions are the one exception: those cannot be rewritten, so a correction that would change one is recorded but not applied.
Accounts and billing
What can I read without an account?
The full earnings history of every company with a public history page, with no ticker limit and no sign-up: what the options market implied before each report, what the stock did, and the measured collapse afterwards. The methodology and the accuracy record are public too.
A free account opens the history of every company we hold, including the ones without a public page yet, and adds the current term structure for a name, the short-straddle outcome tables, and the board for companies reporting today with every number shown (later reports show only the name and date). It costs nothing and nothing about it expires.
What does a paid plan add?
The forward board, which lists every company reporting in the next two weeks and ranks each one with a clean reading by how its report is priced against its own history, plus CSV export and an API key.
The wall is a date, not a quota. Everything that has already happened is free for every name; what costs money is the fortnight that has not happened yet.
What does it cost?
$290/yr, or $29/mo. It is one product either way: the annual plan is the same thing at a lower rate, not a bigger tier.
Can I cancel, and do you refund?
If it is not what you expected, email us within 1 day of your first payment on a plan and we will refund it in full.
Cancelling later stops the next payment and keeps your access to the end of the period you have already paid for. You can cancel yourself from your account page; there is nothing to email us about.
Is there CSV export and an API?
Both, on a paid plan. CSV covers the board and any company’s report history. The API is deliberately small: the board, a company’s reports, and its daily snapshots. You issue your own key from your account page.
Do I need a password?
No. Signing in sends a link to your email address and clicking it signs you in, so there is no password to choose, store or lose.
Using the numbers elsewhere
Can I republish these numbers in my own product?
Not without asking us first. What you read here is derived from market data we license, and the terms you agree to by using the site cover personal and internal use, not redistribution.
If you want to use the numbers in something you publish or sell, email us and say what you have in mind. We are happy to talk, and the answer depends on which numbers and how many.
Can I cite a number from here?
Yes. Quote it, say it came from us, and link the page you took it from. Prices and the accuracy record change, so a link stays current where a copied figure goes out of date.